Skip to content
Core

Planning

Paw plans before it acts, and shows you the plan.

Overview

Every non-trivial request goes through a visible planning stage. Paw breaks work into a live TODO list, tracks progress against it in the Coding Canvas, and adapts the plan honestly when reality doesn't match expectations — rather than pretending a task is done when it isn't.

How it works

  1. 1

    Break down the request

    Paw turns a request into concrete, ordered steps before starting.

  2. 2

    Track live

    Each step's status updates in real time in the Coding Canvas.

  3. 3

    Adapt honestly

    If a step reveals new information, the plan updates — it isn't silently forced to match the original guess.

Use cases

  • • Multi-step engineering tasks that need visible progress, not a black box
  • • Long-running research where you want to see what's been checked so far

Limitations

  • • Planning quality depends on the underlying reasoning model configured for your tier